1. Field of the Invention
The present invention relates to the general field of Integrated Circuit (IC) devices and fabrication methods, and more particularly to multilayer or Three Dimensional Integrated Circuit (3D IC) devices and fabrication methods.
2. Discussion of Background Art
Semiconductor manufacturing is known to improve device density in an exponential manner over time, but such improvements come with a price. The mask set cost required for each new process technology has also been increasing exponentially. While 20 years ago a mask set cost less than $20,000, it is now quite common to be charged more than $1M for today's state of the art device mask set.
These changes represent an increasing challenge primarily to custom products, which tend to target smaller volume and less diverse markets therefore making the increased cost of product development very hard to accommodate.
Custom Integrated Circuits can be segmented into two groups. The first group includes devices that have all their layers custom made. The second group includes devices that have at least some generic layers used across different custom products. Well-known examples of the second kind are Gate Arrays, which use generic layers for all layers up to a contact layer that couples the silicon devices to the metal conductors, and Field Programmable Gate Array (FPGA) devices where all the layers are generic. The generic layers in such devices are mostly a repeating pattern structure, called a Master Slice, in an array form.
The logic array technology is based on a generic fabric that is customized for a specific design during the customization stage. For an FPGA the customization is done through programming by electrical signals. For Gate Arrays, which in their modern form are sometimes called Structured Application Specific Integrated Circuits (or Structured ASICs), the customization is by at least one custom layer, which might be done with Direct Write eBeam or with a custom mask. As designs tend to be highly variable in the amount of logic and memory and type of input & output (I/O) each one needs, vendors of logic arrays create product families, each product having a different number of Master Slices covering a range of logic, memory size and I/O options. Yet, it is always a challenge to come up with minimum set of Master Slices that will provide a good fit for the maximal number of designs because it is quite costly if a dedicated mask set is required for each product.
U.S. Pat. No. 4,733,288 issued to Sato in March 1988 (“Sato”), discloses a method “to provide a gate-array LSI chip which can be cut into a plurality of chips, each of the chips having a desired size and a desired number of gates in accordance with a circuit design.” The references cited in Sato present a few alternative methods to utilize a generic structure for different sizes of custom devices.
The array structure fits the objective of variable sizing. The difficulty to provide variable-sized array structure devices is due to the need of providing I/O cells and associated pads to connect the device to the package. To overcome this limitation Sato suggests a method where I/O could be constructed from the transistors that are also used for the general logic gates. Anderson also suggested a similar approach. U.S. Pat. No. 5,217,916 issued to Anderson et al. on Jun. 8, 1993, discloses a borderless configurable gate array free of predefined boundaries using transistor gate cells, of the same type of cells used for logic, to serve the input and output function. Accordingly, the input and output functions may be placed to surround the logic array sized for the specific application. This method places a severe limitation on the I/O cell to use the same type of transistors as used for the logic and; hence, would not allow the use of higher operating voltages for the I/O.
U.S. Pat. No. 7,105,871 issued to Or-Bach et al. on Sep. 12, 2006, discloses a semiconductor device that includes a borderless logic array and area I/Os. The logic array may comprise a repeating core, and at least one of the area I/Os may be a configurable I/O.
In the past it was reasonable to design an I/O cell that could be configured to the various needs of most customers. The ever increasing need of higher data transfer rate in and out of the device drove the development of special serial I/O circuits called SerDes (Serializer/Deserializer) transceivers. These circuits are complex and require a far larger silicon area than conventional I/Os. Consequently, the variations needed are combinations of various amounts of logic, various amounts and types of memories, and various amounts and types of I/O. This implies that even the use of the borderless logic array of the prior art will still require multiple expensive mask sets.
The most common FPGAs in the market today are based on Static Random Access Memory (SRAM) as the programming element. Floating-Gate Flash programmable elements are also utilized to some extent. Less commonly, FPGAs use an antifuse as the programming element. The first generation of antifuse FPGAs used antifuses that were built directly in contact with the silicon substrate itself. The second generation moved the antifuse to the metal layers to utilize what is called the Metal to Metal Antifuse. These antifuses function like programmable vias. However, unlike vias that are made with the same metal that is used for the interconnection, these antifuses generally use amorphous silicon and some additional interface layers. While in theory antifuse technology could support a higher density than SRAM, the SRAM FPGAs are dominating the market today. In fact, it seems that no one is advancing Antifuse FPGA devices anymore. One of the severe disadvantages of antifuse technology has been their lack of re-programmability. Another disadvantage has been the special silicon manufacturing process required for the antifuse technology which results in extra development costs and the associated time lag with respect to baseline IC technology scaling.
The general disadvantage of common FPGA technologies is their relatively poor use of silicon area. While the end customer only cares to have the device perform his desired function, the need to program the FPGA to any function requires the use of a very significant portion of the silicon area for the programming and programming check functions.
Some embodiments of the present invention seek to overcome the prior-art limitations and provide some additional benefits by making use of special types of transistors that are fabricated above or below the antifuse configurable interconnect circuits and thereby allow far better use of the silicon area.
One type of such transistors is commonly known in the art as Thin Film Transistors or TFT. Thin Film Transistors has been proposed and used for over three decades. One of the better-known usages has been for displays where the TFT are fabricated on top of the glass used for the display. Other type of transistors that could be fabricated above the antifuse configurable interconnect circuits are called Vacuum Field Effect Transistor (FET) and was introduced three decades ago such as in U.S. Pat. No. 4,721,885.
Other techniques could also be used such as employing Silicon On Insulator (SOI) technology. In U.S. Pat. Nos. 6,355,501 and 6,821,826, both assigned to IBM, a multilayer three-dimensional Complementary Metal-Oxide-Semiconductor (CMOS) Integrated Circuit is proposed. It suggests bonding an additional thin SOI wafer on top of another SOI wafer forming an integrated circuit on top of another integrated circuit and connecting them by the use of a through-silicon-via, or thru layer via (TLV). Substrate supplier Soitec SA, of Bernin, France is now offering a technology for stacking of a thin layer of a processed wafer on top of a base wafer.
Integrating top layer transistors above an insulation layer is not common in an IC because the quality and density of prior art top layer transistors are inferior to those formed in the base (or substrate) layer. The substrate may be formed of mono-crystalline silicon and may be ideal for producing high density and high quality transistors, and hence preferable. There are some applications where it has been suggested to build memory bit cells using such transistors as in U.S. Pat. Nos. 6,815,781, 7,446,563 and a portion of an SRAM based FPGA such as in U.S. Pat. Nos. 6,515,511 and 7,265,421.
Embodiments of the present invention seek to take advantage of the top layer transistor to provide a much higher density antifuse-based programmable logic. An additional advantage for such use will be the option to further reduce cost in high volume production by utilizing custom mask(s) to replace the antifuse function, thereby eliminating the top layer(s) anti-fuse programming logic altogether.
Additionally some embodiments of the present invention may provide innovative alternatives for multi-layer 3D IC technology. As on-chip interconnects are becoming the limiting factor for performance and power enhancement with device scaling, 3D IC may be an important technology for future generations of ICs. Currently the only viable technology for 3D IC is to finish the IC by the use of Through-Silicon-Via (TSV). The problem with TSVs is that they are relatively large (a few microns each in area) and therefore may lead to highly limited vertical connectivity. The present invention may provide multiple alternatives for 3D IC with an order of magnitude improvement in vertical connectivity.
Constructing future 3D ICs will require new architectures and new ways of thinking. In particular, yield and reliability of extremely complex three dimensional systems will have to be addressed, particularly given the yield and reliability difficulties encountered in building complex Application Specific Integrated Circuits (ASIC) of recent deep submicron process generations.
Fortunately, current testing techniques will likely prove applicable to 3D IC manufacturing, though they will be applied in very different ways. FIG. 116 illustrates a prior art set scan architecture in a 2D IC ASIC 11600. The ASIC functionality is present in logic clouds 11620, 11622, 11624 and 11626 which are interspersed with sequential cells like, for example, pluralities of flip-flops indicated at 11612, 11614 and 11616. The ASIC 11600 also has input pads 11630 and output pads 11640. The flip-flops are typically provided with circuitry to allow them to function as a shift register in a test mode. In FIG. 116 the flip-flops form a scan register chain where pluralities of flip-flops 11612, 11614 and 11616 are coupled together in series with Scan Test Controller 11610. One scan chain is shown in FIG. 116, but in a practical design comprising millions of flip-flops, many sub-chains will be used.
In the test architecture of FIG. 116, test vectors are shifted into the scan chain in a test mode. Then the part is placed into operating mode for one or more clock cycles, after which the contents of the flip-flops are shifted out and compared with the expected results. This may provide an excellent way to isolate errors and diagnose problems, though the number of test vectors in a practical design can be very large and an external tester may be utilized.
FIG. 117 shows a prior art boundary scan architecture in exemplary ASIC 11700. The part functionality is shown in logic function block 11710. The part also has a variety of input/output cells 11720, each comprising a bond pad 11722, an input buffer 11724, and a tri-state output buffer 11726. Boundary Scan Register Chains 11732 and 11734 are shown coupled in series with Scan Test Control block 11730. This architecture operates in a similar manner as the set scan architecture of FIG. 116. Test vectors are shifted in, the part is clocked, and the results are then shifted out to compare with expected results. Typically, set scan and boundary scan are used together in the same ASIC to provide complete test coverage.
FIG. 118 shows a prior art Built-In Self Test (BIST) architecture for testing a logic block 11800 which comprises a core block function 11810 (what is being tested), inputs 11812, outputs 11814, a BIST Controller 11820, an input Linear Feedback Shift Register (LFSR) 11822, and an output Cyclical Redundancy Check (CRC) circuit 11824. Under control of BIST Controller 11820, LFSR 11822 and CRC 11824 are seeded (i.e., set to a known starting value), the block 11800 is clocked a predetermined number of times with LFSR 11822 presenting pseudo-random test vectors to the inputs of Block Function 11810 and CRC 11824 monitoring the outputs of Block Function 11810. After the predetermined number of clocks, the contents of CRC 11824 are compared to the expected value (or signature). If the signature matches, block 11800 passes the test and is deemed good. This sort of testing is good for fast “go” or “no go” testing as it is self-contained to the block being tested and does not require storing a large number of test vectors or use of an external tester. BIST, set scan, and boundary scan techniques are often combined in complementary ways on the same ASIC. A detailed discussion of the theory of LSFRs and CRCs can be found in Digital Systems Testing and Testable Design, by Abramovici, Breuer and Friedman, Computer Science Press, 1990, pp 432-447.
Another prior art technique that is applicable to the yield and reliability of 3D ICs is Triple Modular Redundancy. This is a technique where the circuitry is instantiated in a design in triplicate and the results are compared. Because two or three of the circuit outputs are always in agreement (as is the case with binary signals) voting circuitry (or majority-of-three or MAJ3) takes that as the result. While primarily a technique used for noise suppression in high reliability or radiation tolerant systems in military, aerospace and space applications, it also can be used as a way of masking errors in faulty circuits since if any two of three replicated circuits are functional the system will behave as if it is fully functional. A discussion of the radiation tolerant aspects of TMR systems, Single Event Effects (SEE), Single Event Upsets (SEU) and Single Event Transients (SET) can be found in U.S. Patent Application Publication 2009/0204933 to Rezgui (“Rezgui”).
Additionally the 3D technology according to some embodiments of the present invention may enable some very innovative IC alternatives with reduced development costs, increased yield, and other important benefits.